my_functions.py

from flask import Flask, render_template, request, send_from_directory
import keras
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense, Dropout, Conv2D, MaxPooling2D, BatchNormalization, Flatten
import numpy as np
import matplotlib.pyplot as plt
from skimage.transform import resize
import urllib.request
import json
import os
import google_streetview.api
import time
import glob
from config import gkey
from datetime import datetime

### ADDRESS INPUT ###
###########returns classifications from user address input###########
def address_form(model, submit_address, ADDRESS_SUBMIT_COUNT):

    #API CALL FOR USER ADDRESS SUBMIT IN FORM
    input_address = []
    # address_count = 0

    # submit_address = request.form["address"]
    input_address.append(submit_address)
    
    address = np.array(input_address)

    np.savetxt("static/data/user_address_submit.txt", address, fmt='%5s')

    #this is the first part of the streetview, url up to the address, this url will return a 600x600px image
    pre="https://maps.googleapis.com/maps/api/streetview?size=600x600&location="
    
    #this is the second part of the streetview url, the text variable below, includes the path to a text file containing one address per line
    #the addresses in this text file will complete the URL needed to return a streetview image and provide the filename of each streetview image
    text="static/data/user_address_submit.txt"
    
    #this is the third part of the url, needed after the address
    suf=f"&key={gkey}&fov=60"
    
    #this is the directory that will store the streetview images
    directory=r"static/images/address_submit/"
    
    #checks if the directory variable (output path) above exists and creates it if it does not
    if not os.path.exists(directory):
        os.makedirs(directory)
    
    #opens the address list text file (from the 'text' variable defined above) in read mode ("r")
    with open(text,"r") as text_file:
        #the variable 'lines' below creates a list of each address line in the source 'text' file
        address_choice = text_file.readline().strip('\n')

        ln = address_choice.replace(" " , "+")
        
        # creates the url that will be passed to the url reader, this creates the full, valid, url that will return a google streetview image for each address in the address text file
        URL = pre+ln+suf
        
            #creates the filename needed to save each address's streetview image locally
        filename = os.path.join(directory, "_" + str(ln)+".jpg")
        
        #final step, fetches and saves the streetview image for each address using the url created in the previous steps
        urllib.request.urlretrieve(URL, filename)
        
        # time.sleep(1)

        renamed_image = f"static/images/address_submit/{ADDRESS_SUBMIT_COUNT}.jpg"

        if os.path.exists(renamed_image):
            os.remove(renamed_image)
        os.rename(filename, renamed_image)
        
    # prepare image for model prediction by first resizing to be consistent with the model
    image = plt.imread(f'static/images/address_submit/{ADDRESS_SUBMIT_COUNT}.jpg')
   
    resized_image = resize(image, (500,500,3))
    data = {}
    predictions = model.predict(np.array([resized_image]))[0]
    predictions = list(predictions)
    best_guess_value = max(predictions)
    best_guess_index = predictions.index(best_guess_value)
    classifications = {0: 'Brick', 1: 'Siding', 2: 'Unknown'}
    best_guess_category = classifications[best_guess_index]

    data['Heading'] = 'THE CLASSIFICATION OF THIS PROPERTY IS:'

    if best_guess_value >= 0.395:
        data['Best_guess'] = best_guess_category + '.'
    else:
        data['Best_guess'] = 'Unknown.'

    for i, prediction in enumerate(predictions):
        data[classifications[i]] = f'{classifications[i]}: {int(round(100*prediction,0))}%'
 
    return (data, predictions, best_guess_category, address)



### IMAGE INPUT ###
###########returns model classifications from user image upload###########

def image_form(model, image):

    resized_image = resize(image, (500,500,3))
    data = {}
    predictions = model.predict(np.array([resized_image]))[0]
    predictions = list(predictions)
    best_guess_value = max(predictions)
    best_guess_index = predictions.index(best_guess_value)
    classifications = {0: 'Brick', 1: 'Siding', 2: 'Unknown'}
    best_guess_category = classifications[best_guess_index]

    data['Heading'] = 'THE CLASSIFICATION OF THIS PROPERTY IS:'

    if best_guess_value >= 0.395:
        data['Best_guess'] = best_guess_category + '.'
    else:
        data['Best_guess'] = 'Unknown.'

    for i, prediction in enumerate(predictions):
        data[classifications[i]] = f'{classifications[i]}: {int(round(100*prediction,0))}%'

    return (data, predictions, best_guess_category)